"value":"In the Linux kernel, the following vulnerability has been resolved:\n\ndm array: fix releasing a faulty array block twice in dm_array_cursor_end\n\nWhen dm_bm_read_lock() fails due to locking or checksum errors, it\nreleases the faulty block implicitly while leaving an invalid output\npointer behind. The caller of dm_bm_read_lock() should not operate on\nthis invalid dm_block pointer, or it will lead to undefined result.\nFor example, the dm_array_cursor incorrectly caches the invalid pointer\non reading a faulty array block, causing a double release in\ndm_array_cursor_end(), then hitting the BUG_ON in dm-bufio cache_put().\n\nReproduce steps:\n\n1. initialize a cache device\n\ndmsetup create cmeta --table \"0 8192 linear /dev/sdc 0\"\ndmsetup create cdata --table \"0 65536 linear /dev/sdc 8192\"\ndmsetup create corig --table \"0 524288 linear /dev/sdc $262144\"\ndd if=/dev/zero of=/dev/mapper/cmeta bs=4k count=1\ndmsetup create cache --table \"0 524288 cache /dev/mapper/cmeta \\\n/dev/mapper/cdata /dev/mapper/corig 128 2 metadata2 writethrough smq 0\"\n\n2. wipe the second array block offline\n\ndmsteup remove cache cmeta cdata corig\nmapping_root=$(dd if=/dev/sdc bs=1c count=8 skip=192 \\\n2>/dev/null | hexdump -e '1/8 \"%u\\n\"')\nablock=$(dd if=/dev/sdc bs=1c count=8 skip=$((4096*mapping_root+2056)) \\\n2>/dev/null | hexdump -e '1/8 \"%u\\n\"')\ndd if=/dev/zero of=/dev/sdc bs=4k count=1 seek=$ablock\n\n3. try reopen the cache device\n\ndmsetup create cmeta --table \"0 8192 linear /dev/sdc 0\"\ndmsetup create cdata --table \"0 65536 linear /dev/sdc 8192\"\ndmsetup create corig --table \"0 524288 linear /dev/sdc $262144\"\ndmsetup create cache --table \"0 524288 cache /dev/mapper/cmeta \\\n/dev/mapper/cdata /dev/mapper/corig 128 2 metadata2 writethrough smq 0\"\n\nKernel logs:\n\n(snip)\ndevice-mapper: array: array_block_check failed: blocknr 0 != wanted 10\ndevice-mapper: block manager: array validator check failed for block 10\ndevice-mapper: array: get_ablock failed\ndevice-mapper: cache metadata: dm_array_cursor_next for mapping failed\n------------[ cut here ]------------\nkernel BUG at drivers/md/dm-bufio.c:638!\n\nFix by setting the cached block pointer to NULL on errors.\n\nIn addition to the reproducer described above, this fix can be\nverified using the \"array_cursor/damaged\" test in dm-unit:\n dm-unit run /pdata/array_cursor/damaged --kernel-dir <KERNEL_DIR>"